The field of federated learning is moving towards addressing the challenges of multimodal data heterogeneity and non-independent and identically distributed (non-IID) data. Researchers are proposing novel frameworks and techniques to enable the co-enhancement of server and client models, seamless blending of horizontal and vertical federated learning, and progressive parameter alignment for personalized federated learning. These advancements have the potential to improve the performance and robustness of federated learning models in real-world settings, such as healthcare and finance, where data privacy is crucial. Noteworthy papers include FedHUG, which proposes a federated heterogeneous unsupervised generalization framework for remote physiological measurements, and BlendFL, which introduces a blended federated learning framework for handling multimodal data heterogeneity. FedPPA is also notable for its progressive parameter alignment approach for personalized federated learning.